httpwww itl waw plczasopismajtit2010140

8/13/2019 Httpwww Itl Waw Plczasopismajtit2010140

1/8

Paper Frequency Offset Compensation

for OFDM Systems Using a Combined

Autocorrelation and Wiener Filtering SchemeAli Ramadan Ali, Tariq Jamil Khanzada, and Abbas Omar

AbstractOne of the orthogonal frequency division multi-

plexing (OFDM) system disadvantages is its sensitivity to fre-

quency offset and phase noise, which lead to losing the or-

thogonality between the subcarriers and thereby degrade the

system performance. In this paper a joint scheme for fre-

quency offset and pilot-based channel estimation is introduced

in which the frequency offset is first estimated using an au-

tocorrelation method, and then is fined further by applying

an iterative phase correction by means of pilot-based Wiener

filtering method. In order to verify the capability of the es-timation algorithm, the scheme has been implemented and

tested using a real measurement system in a multipath indoor

environment. The results show the algorithm capability of

compensating for the frequency offset with different transmis-

sion and channel conditions.

Keywordschannel characterization, frequency offset, OFDM

measurement, Wiener filtering.

1. Introduction

Orthogonal frequency division multiplexing (OFDM) di-vides the base bandwidth into N orthogonal narrow sub-

channels transmitted in parallel. The generation of the

OFDM signals at the transmitter is accomplished using

inverse fast Fourier transform (IFFT), which delivers or-

thogonal carriers. Due to its long symbol duration to-

gether with adding a suitable guard interval (GI) to the

time domain signal, OFDM system can handel the fre-

quency selective fading resulting from the multipath effect.

The source bit stream in OFDM system is first encoded,

interleaved and then mapped using a common modulation

scheme like quadrature amplitude modulation (QAM) or

phase shift keying (PSK).

The resulting symbols are arranged in blocks with length N,

and then IFFT is applied, so that, each symbol is given an

orthogonal carrier. In order to cope with the multipath

effect a suitable GI (longer than the channel impulse re-

sponse) is inserted in the time domain signal before trans-

mission. In OFDM system usually a so called cyclic pre-

fix GI is used, which inserts the last portion of the OFDM

symbol at the beginning. This type of GI allows for cir-

cular convolution with the channel and maintains the or-

thogonality of the carriers. The ability of combating the

frequency selective channels has resulted in choosing the

OFDM system as a standard modulation scheme for trans-

mitting high rate services as in IEEE 802.11a/g wireless

local area network (WLAN) [1], digital audio broadcast-

ing DAB [2], digital video broadcasting (DVB) [3], and

became a promising candidate for fourth generation (4G)

mobile communication systems [4], [5]. On the other hand

OFDM is sensitive to phase noise and frequency offsets,

which cause phase rotation and loss in the orthogonality

between the subcarriers.

This introduces intercarrier interference (ICI) [6][9] as an-

other source of noise, which degrades the system perfor-

mance. The notation to be adopted throughout this paper

conforms to the following convention: bold uppercase let-

tersH denote matrices, bold uppercase underlined letters V

denote frequency domain vectors, while time domain vec-

tors are represented by bold lowercase underlined letters v,

variables in time are denoted by lowercase letters v(n), andin frequency by uppercase letters V(n). The operator{.}H,represents Hermitian transpose. I denotes the identity

matrix.

2. System Model and the Measurement

Setup

Making use of the FFT operation, the time domain trans-

mitted OFDM symbol can be written as [10]

x(n) =N1

m=0

S(m)ej2nm

N 0nN1 , (1)

where S(m) is the QAM symbol at the mth subcarrier. Be-cause of the cyclic prefix, the transmitted signal is circularly

convolved with the impulse response of the multipath time-

variant channel. The received time domain OFDM symbol

is expressed as

y(n) =L1

l=0

h(n, l)x(n l) + v(n), (2)

where h(n, l) and l are the complex random variable andthe corresponding tap delay, respectively, at the lth path

of the multipath channel, and v(n) represents the additivewhite Gaussian noise at sample instant n. After apply-

ing FFT at the receiver side, the frequency domain OFDM

symbol becomes

Y(k) = 1

N

N1

n=0 L1

l=0h(n, l)x(n l) + v(n)e

j 2nkN . (3)

40


2/8

Frequency Offset Compensation for OFDM Systems Using a Combined Autocorrelation and Wiener Filtering Scheme

Equation (3) can be simplified [9] to

Y(k) =S(k)Hk,k+ m=k

S(m)Hk,m

ICI

+V(k), (4)

where:

Hk,m= 1

N

N1

n=0

L1

l=0

h(n, l)ej2N(k(ln)+nk) . (5)

Writing (4) in vector form representation gives

Y =HS + V , (6)

thus the main diagonal of the matrixH represents the chan-

nel response at the subcarriers, while the off-diagonals rep-

resent the ICI components result from the frequency offsets

and the time variation of the channel.

Fig. 1. The measurement system setup.

The measurement setup of the OFDM system is shown in

Fig. 1. As shown from the figure, the measurement system

is a combination of software and hardware components.

2.1. Software Setup

At an external PC, the baseband IQ (in-phase and quadra-

ture-phase) OFDM signal based on IEEE 802.11g standard

is generated using Matlab code. The OFDM symbols are

arranged in a few frames to be transmitted. At the begin-

ning of each frame, some training symbols (preambles) are

inserted, which are used for packet detection and frequency

and time offset estimation. The time domain baseband sig-

nal is sent via LAN connection to the transmitter, in order

to be up-converted and transmitted through the wireless

channel. The commands for controlling the power and the

carrier frequency are sent to the transmitter before sending

the baseband signal. On the other hand, the IQ data from

the captured signal at the receiver is sent back to the ex-

ternal PC via the LAN for processing. Packet detection,

frequency and time offset estimation and pilot-based chan-

nel estimation are all accomplished using Matlab.

2.2. Hardware Setup

2.2.1. The Transmitter

For transmitting the real time multicarrier signal, an MXG

signal generator N5182A [11] has been used. This signal

generator works on a frequency range between 100 kHz

to 6 GHz and transmission power up to 17 dBm. The

baseband IQ data is generated using Matlab and sent to thesignal generator via LAN connection. The signal generator

saves the IQ data, applies the IF modulation and RF up-

conversion and then transmits the signal with predefined

power and carrier frequency. In order to have a continuous

transmission, the signal generator sends the same signal

repeatedly.

2.2.2. The Antennas

The transmit antenna works between 2.4 GHz and 2.6 GHz

with gain 9 dBi at the resonance frequency [12], while

the receive antenna has an ultra-wideband characteristics in

the frequency range from 800 MHz to 18 GHz [13]. Both

antennas have been fabricated in the Institute of Electronics,

Signal Processing and Communications (IESK) at the Otto

von Guericke University of Magdeburg, Germany.

2.2.3. The Receiver

The EXA spectrum analyzer from Agilent [14] has been

used as a receiver. This device has an absolute sensitivity of

79.4dBm, a dynamic range of 93.1 dB, and a maximuminput power of +30 dBm. The reviver down converts the

RF signal and generates the IQ baseband data which is sent

back to the external PC for processing.

3. Wireless Channel Characterization

Estimating the channel parameters such as the number of

multipaths, the delay spread, and the Doppler spread allows

for suitable design and robust behavior of the system. In

this section some measurements have been made in order to

estimate the channel taps and the delay spread of the chan-

nel. An OFDM signal has been sent through the channel,

captured at the receiver and then processed in order to cal-

culate the channel transfer function (CTF) and the channel

impulse response (CIR). The measurements have been con-

ducted in Lab312, which is one of the IESK laboratories at

the University of Magdeburg. The lab was clustered with

furniture, test and measurement equipments. The topology

of the lab and the test scenario is shown in Fig. 2.

During the measurements, the transmit antenna remained

fixed at the same place with different orientations, while the

receive antenna was moved to different places and orienta-

tions. The signal used for measurement was a WLAN sig-

nal with 20 MHz bandwidth and carrier frequency

of 2.412 GHz. The channel transfer function was calcu-

lated as the ratio between the transmitted frequency do-

main OFDM symbols by the received frequency domain

41


3/8


4/8


4. A Joint Frequency Offset and Wiener

Filtering Channel Estimation Scheme

The time and frequency synchronization play an important

role in designing and implementing the practical systems.

In WLAN system, the signal is headed with a preamble at

the beginning of each frame. This preamble is basically

used for time synchronization, frequency offset estimation

and for the initial channel estimation. The preamble con-

tains two parts, the first one is called short training symbols

(STS), which contains 10 short symbols occupy 8 s. The

STS are used for time synchronization and course frequency

offset (CFO) estimation. The second part is called long

training symbols (LTS), which contains two long symbols

each with time duration of 3.2s as well as a GI with time

duration of 1.6 s. The LTS are used for further enhanc-

ing the time synchronization and for fine frequency offset

(FFO) estimation as well as initial channel estimation.

The OFDM symbol of the WLAN system contains 64 car-

riers, 52 of them are active carriers, while the other 12 car-riers work as frequency guard at both sides of the spectrum.

For channel estimation and equalization proposes, 4 carri-

ers out of the 52 are booked for pilots.

In order to start processing the received OFDM signal, the

signal packets need to be detected first and the first sam-

ple of the OFDM symbols needs to be indicated in order

to apply the FFT operation on the right symbol samples.

A suitable method for detecting the packet is based on the

autocorrelation of the received preamble [16], which con-

tains short and long training symbols. The auto correlation

is written in the following form

Z(n) =NT S1

c=0

y(n + c)y(n + c +NTS), (10)

wherey(n)represents thenth incoming sample of the base-band signal, y(n) represents the conjugate of y(n) andNTS= 16 samples in case of short training symbols. Inorder to avoid the expected variance of the incoming sig-

nal power, the autocorrelation needs to be normalized by

a moving sum of the received signal power [17]. The mov-

ing sum of the received power can be written as

P(n) =

NTS1

c=0

|y(n + c +NTS)|2. (11)

The normalized autocorrelation used for packet detection

reads

M(n) = |Z(n)|2

P(n) . (12)

The packet is detected at the first peak ofM(n)after a pre-defined threshold, which should be chosen to minimize the

possibility of false alarm. Using only the short training

symbols in packet detection gives a reasonable performance

in case of good conditions of the channel, however, be-

cause of the expected time offsets at bad channel conditions,

it is better to use both short and long training symbols for

this purpose.

Figure 5 shows the autocorrelation function M(n), whosefirst peak indicates the beginning of the packet (the first

sample of the short training symbols), while the second

Fig. 5. The packet detection based on autocorrelation of the

preamble.

peak indicates the first sample of the GI of the long training

symbols. By taking the difference between the two peaks,

the time offset, if found, can be estimated and then the start

of the packet is corrected.

Figure 6 presents a comparison between the spectrum of

the transmitted and the received OFDM signals. The figure

shows the effect of the widowing at the receiver side, which

reduces the side loops of the spectrum.

Fig. 6. A comparision between the transmitted and measured

spectrum of the OFDM signal.

As a consequence of the frequency offsets between the os-

cillators of the transmitter and the receiver, phase errors

and ICI occur on the received signal and lead to degrad-

ing the OFDM system performance. The frequency offset

can be estimated from the autocorrelation of the training

symbols in the preamble. If there is a frequency offset fe,

43


5/8

Ali Ramadan Ali, Tariq Jamil Khanzada, and Abbas Omar

the autocorrelation function will be modified by the phase

error as

Z(n) =ej2feNT SNTS1

c=0

|y(n + c)|2. (13)

The phase introduced by the frequency offset is calculated

as

= Z(n).

The course frequency offset CFO is calculated in case of

the short symbols as

fe=

216T, (14)

where T is the sampling time which is (50 ns) for WLAN

system. For better frequency offset estimation, the esti-

mation is accomplished in two stages CFO and then FFO

estimations. After the compensation of the frequency off-

set, a Wiener filtering method described is applied to fur-

ther correcting for the phase rotation, and at the same time

equalizing the OFDM symbols. In order to enhance theperformance of the frequency offset compensation, a mini-

mum mean square error (MMSE) [18][20] estimator based

on Wiener filtering has been used. The frequency domain

MMSE estimator minimizes the following formula:

E{|e|2}=E{|Hp Hp|2} .The frequency domain channel matrix can be estimated by

linearly filtering the received signal for each channel ele-

ment. The estimated element(m,k) of the channel matrixis expressed as follows:

Hm,k= WH

m,kY, (15)

where Wm,k is the Wiener filter vector that is used to esti-

mate the channel matrix element (m,k).The Wiener filter for each channel element is designed and

optimized according to the MMSE criterion as follows:

W(opt)m,k =arg min

W

EHm,kWHm,kY2. (16)

The solution of Eq. (16) is given by [21]

W(opt)m,k =R

1Pm,k, (17)

where R =E{YYH} is the autocorrelation of the receivedfrequency domain signal andPm,k=E{H

m,kY}is the cross

correlation between the received signal and the actual fre-

quency response of the channel. The derivation of these

correlation matrices is detailed in Appendix.

In order to correct the phase rotation of the captured OFDM

symbols, an iterative scheme is applied to find the phase

error that minimizes the following cost function:

MSE=E{|Sp Sp|2} , (18)

where Sp andSp are vectors contain the transmitted and

equalized pilots.

Fig. 7. The constellation of the equalized 16 QAM modulated

OFDM symbols with different transmission power: (a) 0 dBm,(b) 5 dBm, (c) 10 dBm, (d) 17 dBm at distance ( d=2 m); gray:received symbols, black: equalized symbols.

Fig. 8. An example of transmitting a gray scaled image through

the wireless channel: (a) the transmitted image; (b) the received

un-equalized image; (c) the equalized image with a transmission

power 10 dBm; (d) the equalized image with a transmissionpower 0 dBm; (e) the constellation of the equalized image with

10dBm power; (f) the constellation of the equalized image with

0 dBm power.

44


6/8


7/8

Ali Ramadan Ali, Tariq Jamil Khanzada, and Abbas Omar

of 17 dBm and variant distance between the transmitter and

the receiver. It can be seen from the figure that in addition

to the path loss, the signal gets attenuated (shadowed) at

certain distances because of the multipath effect.

5. Conclusion

This paper presents a method for estimating the wireless

channel for OFDM system. A measurement system based

on off-line measurement has been used for testing the

OFDM signal. This system has been used to characterize

the multipath channel and test the transmission, reception

and channel estimation of the OFDM-based WLAN system.

A method of joint channel and frequency offset estimation

based on the signal preamble and Wiener filtering has been

implemented. The measured results show that, combining

the Wiener filtering and the autocorrelation can compen-

sate for ICI and the phase rotations caused by frequency

offsets and multipath effect.

Appendix

Calculating R and P

Assuming that H, S, and V are uncorrelated and making

use of Eq. (4), the element(k,k) ofR can be written as

Rk,k = E{Y(k)Y(k)}

= EN1

m=0

Hm,kS(m) N1

m=0

Hm,kS(m)

+E{V(k)V(k)} . (21)

From Eq. (21),R can be separated into two parts, the pure

correlation matrix and the noise

R =R +2v I , (22)

where2v is the noise variance, which is the first statistical

parameter that needs to be estimated at the receiver. Also

the element k ofPm,kis calculated as

Pk

m,k= E{Y(k)Hm,k}

= EN1

m=0

Hm,kHm,kS(m

)+E{V(k)Hm,k} . (23)

Assuming a wide sense stationary uncorrelated scattering

(WSSUS) channel model, Rk,k and P

k

m,k can be given by

the following equations [22]:

Rk,k = 2s

m/Sp

L1

l=0

l,v(km,km)dv

+ m,mSp

S(m)S(m)L1

l=0

ej2(mm)fl

l,v(km,km)dv ; (24)

Pk

m,k= mSp

S(m)L1

l=0

ej2(mm)fl

l,v(km,km)dv , (25)

where

l,v(a,b) =sinc(vTs a)sinc(vTs b)Dl(v) ,

where sinc() is the sinc function and Dl(v) is the Dopplerspectrum for each tap. This term can be ignored in the case

of time-invariant channels.

References

[1] C. Smith and J. Meyer, 3G Wireless with 802.16 and 802.11. New

York: McGraw-Hill, 2004.

[2] Radio Broadcasting Systems; Digital Audio Broadcasting (DAB)

to mobile, portable and fixed receivers, ETSI EN 300 401 V1.4.1,

2006.

[3] Digital Video Broadcasting (DVB); Interaction channel for Digital

Terrestrial Television (RCT) incorporating Multiple Access OFDM,

ETSI EN 301 958 V1.1.1, 2002.

[4] P. Lescuyer and T. Lucidarme, Evolved Packet System (EPS): The

LTE and SAE Evolution of 3G UMTS. Chichester: Wiley, 2008.

[5] S. Hara and R. Prasad, Multicarrier Techniques for 4G Mobile Com-

munications. Boston, London: Artech House, 2003.

[6] J. Linartz and A. Grokhov, New equalization approach for

OFDM over dispersive and rapidly time varying channel, in Proc.

PIMRC 2000 Conf., London, UK, 2000.

[7] Y. Liao and K. Chen, Estimation of stationary phase noise by the

autocorrelation of the ICI weighting function in OFDM systems,

IEEE Trans. Wirel. Commun., vol. 5, no. 12, pp. 33703374, 2006.

[8] M. Chang, A novel algorithm of interference self-cancellation for

OFDM systems, IEEE Trans. Wirel. Commun., vol. 6, no. 8,

pp. 28812893, 2007.

[9] W. Jeon, K. Chang, and Y. Cho, An equalization technique for

orthogonal frequency-division multiplexing systems in time-variant

multipath channels, IEEE Trans. Wiurel. Commun., vol. 47, no. 1,

pp. 2732, 1999.

[10] R. Nee and R. Prasad, OFDM for Wireless Multimedia Communi-

cations. Boston, London: Artech House, 2000.

[11] Agilent N5182A MXG Vector Signal Generator, Agilent Technol.,

2007, Data sheet, ftp://ftp.testequity.com/pdf/n5182a.pdf

[12] D. Mittelstrasz, Design Einer Antenne Bestehend aus Primaer-

strahler und Dielektrischer Linse fuer das ISM-Band 2.4 GHz, Stu-

dienarbeit: Otto-von-Guericke Universitaet Magdeburg, 2007.

[13] A. Bandyopadhyay, M. Anis, A. Joestingmeier, T. Meyer, and

A. Omar, Analysis of a compact ultra-wideband antenna for radio

frequency applications, in Proc. Antenn. Propag. Soc. Int. Symp.,

Honolulu, Hawaii, USA, 2007, vol. 9, pp. 19651968.

[14] Agilent EXA Signal Analyzer N9010A, Agilent Technol., 2009,

Data Sheet,

http://cp.literature.agilent.com/litweb/pdf/5989-6529EN.pdf

[15] T. Rappaport, Wireless Communications: Principles and Practice.

New York: Prentice Hall, 2001.

[16] J. Heiskala and J. Terry, OFDM Wireless LANs: A Theoretical and

Practical Guide. New York: Sams, 2001.

[17] T. Schmidl and D. Cox, Robust frequency and timing synchro-

nization for OFDM, IEEE Trans. Commun., vol. 45, no. 12,

pp. 16131621, 1997.

46


8/8


[18] H. Van Trees,Detection, Estimation, and Modulation Theory. Part I:

Detection, Estimation, and Linear Modulation Theory. New York:

Wiley, 1968.

[19] H. Poor, An Introduction to Signal Detection and Estimation. Berlin:

Springer, 1994.

[20] S. Kay,Fundamentals of Statistical Processing. Volume I:Estimation

Theory. Signal Processing Series. Upper Saddle River: Prentice Hall,

1993.

[21] M. Barkat, Signal Detection and Estimation. Norwood: ArtechHouse, 2005.

[22] C. Sgraja and J. Linder, Estimation of rapid time-variant channels

for OFDM using wiener filtering, IEEE Proc. Commun., vol. 4,

pp. 23902395, May 2003.

Ali Ramadan Ali received the

B.Sc. degree in electrical en-

gineering from Altahadi Uni-

versiy, Sirt, Libya, in 1997

and the M.Sc. degree in elec-

trical engineering from Darm-

stadt University of Applied Sci-

ence, Germany, in 2004. He

works toward the Ph.D. de-

gree at the Otto-von-Guericke

University of Magdeburg, Ger-

many. His current research field covers the areas of digital

signal processing for OFDM systems and microwave filter

design. He authored or co-authored over 23 technical pa-

pers in the field of wireless communications and microwave

filters.

e-mail: [email protected]

Otto-von-Guericke University of Magdeburg

Institute for Electronic, Signal Processingand Communications (IESK)

Chair of Microwave and Communication Engineering (HF)

P.O. Box 4120

Universitaetsplatz 2

39016 Magdeburg, Germany

Tariq Jamil Khanzada re-

ceived the M.Eng. degree in

communication systems and

networking, and B.Eng. degree

in computer systems from

Mehran University of Engineer-ing and Technology (MUET),

Pakistan, in 2004 and 1999, re-

spectively. He currently works

at the Institute of Electronics,

Signal Processing and Com-

munications (IESK), Otto-von-Guericke University of Mag-

deburg, Germany, since June 2005 in order to pursue his

Ph.D. degree. Prior to start his Ph.D. studies he has been

working as an Assistant Professor at the Department of

Computer Systems and Software Engineering at MUET,

since 2004 and as a lecturer since 2001. He is also a co-

author of a practical book on computer workshop pub-

lished in 2004. He is also an author of some internationalresearch publications.



Institute for Electronic, Signal Processing

and Communications (IESK)


P.O. Box 4120



Abbas Omar received the

B.Sc. and M.Sc. degrees inelectrical engineering from Ain

Shams University, Cairo, Egypt,

in 1978 and 1982, respectively,

and the Ph.D. degree in electri-

cal engineering from the Tech-

nical University of Hamburg,

Germany, in 1986. Since 1990,

he has been a Professor of elec-

trical engineering, and since

1998, a Director of the chair of microwave and commu-

nication engineering with Otto-von-Guericke University of

Magdeburg, Germany. He authored or co-authored over250 technical papers extending over a wide spectrum of

research areas. His current research fields cover the areas

of microwave, nuclear magnetic resonance, and ultrasonic

imaging, remote sensing, microwave measurements, indoor

and outdoor positioning systems, wideband wireless (ter-

restrial and mobile) communication, subsurface tomogra-

phy and ground penetrating radar, ultra-wideband anten-

nas, and field theoretical modeling of microwave systems

and components. He is a member of the Editorial Board

of some international technical periodicals.



Institute for Electronic, Signal Processingand Communications (IESK)


P.O. Box 4120



47

httpwww itl waw plczasopismajtit2010140

Documents